平面交叉口事故预测及突出事故形态鉴别  

Prediction of Accident at Level Crossing and Distinguishing of Protruding Accident's Formation

作　　者：冯丽生[1] 孟祥海[2] 吕悦晶[2] 

机构地区：[1]黑龙江省收费公路管理局,哈尔滨150010 [2]哈尔滨工业大学交通科学与工程学院,哈尔滨150090

出　　处：《北方交通》2009年第5期127-130,共4页Northern Communications

摘　　要：应用神经网络技术建立了基于三层感知机的平面交叉口事故预测模型,该模型的特点是可预测出具体事故形态的事故数。模型检验结果表明,预测精度达到了89%。在得到分事故形态的事故数预测结果后,基于"突出性"原理建立了一种鉴别事故多发形态的方法。The model for accidents prediction at the level crossing, based on the triple perception, is established applying the neural net techniques. Present model is characterized enabling to predict number of accident formation. Results of model inspection indicates the accuracy of prediction amouts to 89 percent. After prediction of formation accidents＇number,there has been accomplished a method for distinguishing the accident formation frequently occurring,based on ＇ protrusion＇ principle.

关 键 词：事故预测 多发事故类型 人工神经网络 三层感知机 突出性原理 

分 类 号：U412.351[交通运输工程—道路与铁道工程]